Abstract
Much energy has been invested in the development of human-like robots and animated figures capable of socially interacting with humans. Mori (1971) suggested that as a robot becomes more human-like, observers will rate it with increasing familiarity until it reaches a point at which subtle differences from human-like appearance and movement produce a feeling of profound discomfort in the observer, an effect known as the Uncanny Valley. An alternative idea, based on theory of motor resonance, suggests that as the movements of a robot become more human-like, they resonate with our own motor representations and we will be more likely to consider them familiar. Here we used a computer generated figure animated with a walking movement captured from a human actor. Movement was varied in its naturalness across 10 levels by manipulating a) the joint articulation of the arms and legs; b) the phase relationship between opposing arms and legs; or c) increasing the magnitude of a short, physiologically implausible jerk at a random point in the walk cycle. Two groups of participants viewed either a mannequin (N=22) or a human-like figure (N=18) walking for 4s, then rated the walker according to humanness, familiarity, and eeriness. Ratings across all three movement manipulations and for both figure types increased (or decreased for eeriness) linearly as the movement became more natural, inconsistent with the Uncanny Valley hypothesis. Plots of familiarity and eeriness ratings as a function of humanness also showed a linear increase (or decrease) with no valley, even as humanness ratings approached the maximum possible rating. These findings are more consistent with the proposal that as a computer-generated figure's movements become more like the movement of a human, they produce increasingly greater feelings of familiarity. Resonance of the figures movements within the motor system of the observer may underlie such feelings.